Skip to content
Rahul Shiyaniworkspace
Contact

Projects

Case studies and shipped builds

This page expands on AI agents, RAG workflows, predictive modeling, and product delivery work. Every card links to a dedicated project detail page.

AI AGENT + RAG SYSTEM

INTELLIGENT-LOAN-APPROVAL-AGENT.MD

Intelligent Loan Approval Agent

A decision-support system for loan workflows that combines multi-step LLM orchestration, retrieval, and structured knowledge grounding.

  • Indexed multi-format files in under one minute to accelerate review setup.
  • Combined LLM reasoning with ML inference for practical decision support and clearer rationale.
PythonStreamlitAnthropic SDKChromaDBMiniLMRAG

HEALTHCARE ML APP

LIVER-CIRRHOSIS-STAGE-PREDICTION.MD

Liver Cirrhosis Stage Prediction

A live machine learning application that predicts liver cirrhosis stage through a simple web interface built for accessible clinical-style risk assessment.

  • Deployed as a live web app so predictions can be tested without local setup.
  • Focused on turning model output into a clean, usable experience rather than leaving the work as notebook-only analysis.
PythonMachine LearningVercelWeb AppPrediction Workflow

FINANCE + EXPLAINABLE ML

RISK-SCORING-FOR-LOAN-APPROVAL.MD

Risk Scoring for Loan Approval

An explainable ML workflow for loan approval that combines structured preprocessing, risk scoring, and business-facing decision support.

  • Improved model performance by 13% while reaching 86% accuracy.
  • Used explainability layers so predictions could support business decisions instead of acting like a black box.
PythonSQLSHAPSMOTELogistic Regression

ANDROID + ML APP

WASTE-RECOGNITION-APP.MD

Waste Recognition App

An Android application that helps users identify waste items and improve disposal decisions through ML-backed recognition and guided flows.

  • Achieved 86% accuracy in waste recognition through an Android-integrated ML workflow.
  • Reduced load times by 40% while improving reliability through app and data-flow refinement.
JavaAndroidFirebaseTensorFlow Lite

NLP + DEEP LEARNING

ADVANCED-SENTIMENT-ANALYSIS-USING-BERT-AND-LSTM.MD

Advanced Sentiment Analysis using BERT & LSTM

A comparative NLP project that fine-tunes BERT and benchmarks it against deep learning baselines for sentiment classification.

  • Achieved 92.6% accuracy with a fine-tuned BERT model.
  • Outperformed 2 baseline architectures by 12% through structured comparative evaluation.
PythonPyTorchTransformersBERTLSTMCNN

NLP + EXPLAINABLE CLASSIFICATION

FAKE-NEWS-CLASSIFICATION-WITH-LLM-TECHNIQUES.MD

Fake News Classification with LLM Techniques

A text-classification project focused on fake/real news detection using multiple NLP representations with explainability and class-balance handling.

  • Built a binary news-classification workflow using feature-based and NLP-driven representations.
  • Added SHAP and SMOTE to improve interpretability and class-balance handling in model evaluation.
PythonNLPScikit-learnTF-IDFWord2VecSHAPSMOTE

DATA ANALYTICS + VISUALIZATION

AIRBNB-STRATEGIC-DATA-ANALYSIS.MD

Airbnb Strategic Data Analysis

A data analytics project focused on pricing, occupancy, and location-driven patterns that translate raw listing data into business insight.

  • Improved market insight depth by 18% through structured pricing and occupancy analysis.
  • Derived 3+ business insights that strengthened decision support by 12%.
PythonSQLPandasFoliumData Visualization

FULL-STACK PRODUCT + HACKATHON

HABIT-TRACKER.MD

Habit Tracker

A habit-tracking product built in a fast-paced hackathon setting with live progress views, real-time data, and a usability-first interface.

  • Won 1st Place at the Conestoga Hackathon.
  • Shipped 3+ core features and improved usability by 15% in a rapid product build.
ReactFirebaseJavaScriptHTMLCSS

EDTECH + INFORMATION PLATFORM

EDUCOOL-GUIDER.MD

EduCool: Guider

An education-focused web platform that combines school discovery, mapped location data, and embedded resources in one guided experience.

  • Integrated school-location discovery with Google Maps-style navigation support.
  • Combined maps, videos, and URLs into a single education guidance workflow.
jQueryPHPBootstrapCSSGoogle Maps APIWordPressScriptFlask

PANDEMIC HACKATHON + DIGITAL COMMERCE

ONLINE-BUSINESS-MART.MD

Online Business Mart

A hackathon-built digital business concept created during the pandemic to explore online commerce and chatbot-assisted workflow support.

  • Built during Hack Against COVID, where the team ranked among the top 42 globally.
  • Explored digital commerce with chatbot-assisted support under fast-moving hackathon constraints.
AndroidProduct PrototypingChatbot ConceptUiPath / RPA Idea

HEALTHCARE + PRODUCT PROTOTYPE

VIRTUAL-HEALTHCARE-ASSISTANT.MD

Virtual Healthcare Assistant

A healthcare-oriented assistant prototype designed to make guidance and health-related support more accessible through an interactive product flow.

  • Focused on improving accessibility to healthcare-oriented guidance through an interactive assistant experience.
  • Framed a user-centric healthcare workflow in a hackathon-style product prototype.
App DevelopmentHealthcare AssistantPrototype DesignInteractive Workflow

CLASSICAL ML + COMPARATIVE ANALYSIS

BREAST-CANCER-AND-HOUSE-VOTES.MD

Breast Cancer and House Votes

A structured machine learning project that compares model behavior across two benchmark tabular datasets with different feature characteristics.

  • Worked across 2 benchmark datasets to compare classification behavior and preprocessing decisions.
  • Strengthened understanding of feature handling, model selection, and tabular ML evaluation.
PythonScikit-learnClassificationData PreprocessingTabular Data